Biotational

-> Previously I used this blog to log the beginning of my journey of figuring out Single Cell RNA sequencing data figure production and analysis, but I stopped logging my progress. I am still working with a professor to analyze her research on Colo-rectal Cancer in mice, but I am not as hands-on now that I graduated.

-> Now I will just use this blog to write about projects I am undertaking while I work a full-time retail job. Right now I'm enamored with flip phones and want to preserve, emulate, and modify their OS's, though it's hard because a lot of older phones use proprietary software.

This post on the XDA forums lead me to the Qualcomm Product Support Tools (QPST), which can be used to extract the boot image and "flashing...firmware on devices running on Qualcomm Chipsets". According to this site, the OJV has a Qualcomm cpu so we should be golden.

Before I do ANYTHING to my phone, I want to test it out on a defective OJV I might be able to procure from my workplace, but that has to wait until tomorrow. I'll continue this later. Any advice or comments would be appreciated.

I got my hands on an Orbic Journey V phone recently and think its OS is atrocious for various reasons that you can find out by doing a cursory google search about this phone. It came out in 2019, which means it works on my mobile carrier and I have a working number on it, but the main issues with flip phones these days is that they're made for older folks so they're bigger than usual (not really an issue whatever) and crazily, LAZILY, rudimentary. I'm not looking for a crazy pimped out phone here, but some stuff on this thing drives me insane. Why is my Pantech PN-300 more charming and why does it have a better UI than this thing? Insane.

Anyway I want to see if I can reverse engineer/modify the OS or UI in any way, at least to correct some of the typos. Here's my general course of action based on research I did (You can click each step to see my post about it):

Extract boot.img

Reverse Engineer the boot.img/identify aspects of the OS

Make changes

Flash it back onto the phone

This is gonna be tough for a few reasons. - I have no experience with flashing software onto a flip phone - the orbic journey v uses a proprietary OS which is a problem because it's not easily modifiable.

HOWEVER, some good news..... - The proprietary OS is AOSP 8.1 which means we're not working with anything that doesn't already have a some forum posts - and the "Fastboot interface is unlocked, so you can flash anything in this mode as you wish." [bananahackers] which means we can make edits in EDL mode.

Okay here we go

[Citations under the read more]

Scanpy is one of the first packages I installed. It was mentioned by my Research Supervisor (RS) and some other reference material I have read that I haven't posted to this blog yet. But Scanpy is used extensively in the bioinfo/comp bio world to create graphs and organize data. My primary goal is to go through Scanpy's tutorials with their default dataset and then analyze the dataset from a paper that one of my colleagues presented to us two Thursdays ago (7/11) (I didn't realize it was 7/11 and I didn't get a free slurpee...). Below is the link to the official Scanpy website's tutorials page. I will post a reference post with this same link, but for convenience I will also post it here:

Preprocessing and Clustering...

Copying and pasting the code included with this section of the tutorial and running it grants me these results, which match the tutorial:

I don't completely understand how it pulled the database it's using, but I'll figure it out once I complete the tutorial.

Quality Control...

Copying and pasting the code included with this section of the tutorial and running it grants me these results, which match the tutorial:

I wanted to play around with the values a bit and see what happened to how the data was displayed. If I change the jitter value in the block of code that creates the violin plots:

to 0.9, this is what my results were:

As expected, it spread out the values a little more. Cool! Ok changing it back.

Doublet Detection and Normalization...

This is where I encountered some problems. In order to detect doublets, scanpy uses the detection method scrublet, which is awesome sauce, but the command was not working for me.

I kept getting an error from the scanpy files in charge of scrublet. I thought maybe I was missing something from the package, since it said skimage is missing, so I reinstalled scanpy but nothing changed. On the tutorials page there's a note that talks about alternative methods for doublet detection, so if I don't find a solution I can resort to installing one of those. But, I'm going to see if anyone's posted about this issue.

Someone on stack overflow suggested to install scikit-image to fix it, so I did and it worked! skimage does sound like it comes from the scikit-image package. I thought it wouldn't work because the console was taking a while to process like it does when it's gonna give me an error, but nope!

No error and it even went ahead and did the normalization no problem. Awesome sauce!

Feature Selection...

Got the same graphs as the tutorial, which means I'm on track. The tutorial says that this function:

annotates the genes by reproducing implementations of the specific methods depending on a chosen flavor. I'm not really sure what that flavor is or how I can pick it. I see a number value though so I'm gonna change it. Here's the graphs when I run this command with n_top_genes set to 5000.

Hmm... It didn't change anything, which is a shame. The program takes a lot of time processing the doublets so each time I execute the script again it takes a few minutes, which is not good for me because I love getting distracted while it's processing.

I'm changing the value back to 2000 and moving on, but definitely something to come back to and investigate.

Dimensionality Reduction...

Also got the same results as the tutorial, though my first graph displayed a little wider than the tutorial's graph. The first graph "gives us information about how many PCs [Principal Components] we should consider in order to compute the neighborhood relations of cells, e.g. used in the clustering function leiden() or tsne()."[1] It's better to overestimate than underestimate how many PCs. The second graph plots these PCs to "see if there are any potentially undesired features (e.g. batch, QC metrics) driving signifigant variation in this dataset."[1] According to the tutorial, nothing stands out as "too alarming", but the check was a good idea just in case there is anything that skews the data.

Nearest Neighbor Graph Construction and Visualization...

The only way for me to see the legend on the left side was if I stretched it out a bunch like this.

On the left is the output from the tutorial website, on the right is my output if I don't stretch it out. It's upside down? Each individual clump is pretty intact, but they're rotated different ways and in different positions than the tutorial reference image. I'm not sure what I did wrong, since I am just copying and pasting the code from the tutorial right now. I'm going to play around with it and see if I can fix anything later.

Clustering...

At first my console gave me an error but when I ran it again it worked just fine, so I'm not exactly sure what happened. But it looks like the tutorial sample image, aside from the clusters being in their new spots as shown by my previous figure. I also figured out how to fix how the graph looks, but I have to manually fix it each time. I'm planning to try to find out what can fix this, or at least what can set my defaults.

Re-assess quality control and cell filtering...

Same thing as above, seems to work just fine other than the fact that the clusters are in a different place from the tutorial examples.

Last week I prepared my GitHub for the python code I will be preparing. I created a new repository and relearned how to push things into it using git. I had done it before for one of my computer engineering design courses but that was in May and it's been a little while. I thought I already wrote myself a .txt little guide but I didn't so I wrote a new one.

(I like to write myself little guides like this so I can easily refer back to them. I did the same thing with quickly creating React Native apps back when that was something I was doing frequently.)

I forgot how to pull repositories from GitHub but that's nothing a quick online search can't remedy when I get to it.

The repository is currently private, mostly because I assume I will be working with unreleased data that I shouldn't release online like that, but I might clone the repo with only my attempts using public datasets to release publicly.

I also learned how to create a virtual environment, which I haven't really needed to do before, but my Research Supervisor (RS) suggested I do it so that my analysis attempts won't conflict with each other.

https://onlinelibrary.wiley.com/share/MGKGGG55MMZTRF8DHIMY?target=10.1002/ctm2.694

-> I will primarily be using Python since I am familiar with it, but I will probably also dip my toes into R as it's another popular coding language in this field.

-> I am very new to the world of bioinformatics and have memory issues, so I am logging my progress here as a resource for both myself and others as I figure it out.